Development of a Methodology for Evaluating the

Reliability of Transformer Differential Protection

Function Based on Monte Carlo Method
J. R. M. S. Souza, C. S. Pereira Filho, A. R. De Conti

Abstract-- The aim of this paper is to present a time domain

methodology to support the setting of transformer differential
protective relays. This methodology, which is implemented in
Scilab and ATP (Alternative Transients Program), uses concepts
related to reliability theory and to Monte Carlo methods. Four
different types of events were considered to illustrate the model
application, namely transformer energization, external faults,
internal bushings faults, and turn-to-ground faults. It is shown
that the crossblocking technique leads to the highest success rate
of the differential protection of a three-phase 41 MVA
138/34.5/13.8 kV power transformer during energization,
compared with the harmonic restraint and independent harmonic
blocking techniques. The sensitivity of the relay adjustments to
the harmonic reference settings is also discussed. For the
simulation of turn-to-ground faults, the obtained results suggest
that the percentage restraint differential technique is not
sufficiently accurate for transformer protection. On the other
hand, a 100% reliability was obtained for the tests with external
and internal bushing faults. The implemented relay model was
validated through comparisons with data obtained from the
operation of an actual relay in the laboratory. This suggests that
the proposed methodology could be a useful tool for setting the
transformer differential protection.
Keywords: Transformer, differential protection, time domain
analysis, reliability, Monte Carlo Method.

I. INTRODUCTION

here are several functions available in digital protective

relays that are related to transient phenomena. Examples
of this are schemes available in transformer differential
protective relays to detect inrush currents, overexcitation, and
current transformer saturation. In spite of this, the setting of
transformer differential protective relays is often based on
frequency domain studies, such as fault and power flow
analysis [1]. Functions related to electromagnetic transients are
usually set based on either manufacturer recommendations,

J. R. M. S. Souza is with COPEL, Division of Subtransmission System

Protection, Curitiba PR, Brazil (joao.soares@copel.com).
C. S. Pereira Filho is with the Federal University of Minas Gerais UFMG, Belo Horizonte MG, Brazil (clever@ufmg.br).
A. R. De Conti is with the Federal University of Minas Gerais - UFMG,
Belo Horizonte MG, Brazil (conti@cpdee.ufmg.br).
Paper submitted to the International Conference on Power Systems
Transients (IPST2013) in Vancouver, Canada July 18-20, 2013.

generic studies made in the past, or staff experience. Although

consolidated, this kind of procedure is limited in the sense that
it does not take into account particularities that in some cases
may compromise the protection system performance.
In this paper, a new time domain methodology is proposed
to support the setting of transformer differential protective
relays. In this methodology, which is implemented in the
Scilab-ATP environment, concepts related to Reliability
Theory and to Monte Carlo methods are used to identify the
effectiveness of the differential protection of power
transformers. To test the proposed methodology, several cases
related to the following types of events were simulated:
transformer energization, external faults, internal bushings
faults, and turn-to-ground faults. The validity of the
implemented model was then verified by means of
comparisons with measurements performed with an actual
relay.
This paper is organized as follows. Section II presents the
basic concepts related to transformer differential protection
and inrush current detection. Section III discusses the concept
of reliability of the protection function, whose evaluation is the
aim of this study. Section IV describes the proposed
methodology, presenting both the modeling of the system
components and the formulation of the Monte Carlo method.
Section V presents results of application of this methodology
to a series of events related to a 41 MVA three-winding
transformer. Section VI illustrates the relay model validity by
comparing the obtained results with the performance of an
actual relay. Finally, Section VII presents the conclusions.
II. TRANSFORMER DIFFERENTIAL PROTECTION
A. Basic Concepts
Differential protection is one of the most popular techniques used for transformer protection because of its simplicity
and efficiency. It is based on Kirchhoffs law, which
enunciates that the sum of all currents flowing into a node is
equal to zero. In other words, the current entering a device
must be equal to the current leaving this device unless an
internal fault occurs [1].
In a device with n branches, a differential current Idiff can be
defined as in (1), where Ij is the current flowing through
branch j. This differential current is also called operation
current [2-4].

B. Inrush Current Detection Techniques

I
I diff =

j =1

(1)

When the protected device is a power transformer, the

transformation ratio and the angular displacement must be
compensated before (1) is calculated by the relay. The
transformation ratio can be compensated by changing the
current transformers ratio and/or by changing internal tap
settings [15]. Angular displacement can be compensated by
current transformer conections or by internal settings such as
compensation matrices (only in digital relays) [4]. However, it
is usually impossible to compensate for all sources of errors.
Some conditions, such as current transformer errors and power
transformer tap changes may produce a false differential
current. To prevent misoperation, the differential current has to
be compared with another current which is proportional to the
current flowing through the power transformer. This current is
called restraint current [2-4].
Each relay manufacturer proposes the use of a different
equation to calculate the restraint current Irest. Some examples
are as follows.
n

I
I rest =

j =1

I rest =

[4]

(2)

[2]

(3)

[3]

(4)

n
n

j =1

( )

I rest = max I j

Differential and restraint currents define an operation

characteristic such as that shown in Fig. 1. In this particular
case, the protection scheme is called percentage restraint
differential.

Inrush currents are created during transformer energization

because of the non-linearity of the transformer core [15]. Such
currents present a rich harmonic content, which can be used
for detecting their occurrence and avoiding the misoperation of
the differential relays designed to protect the power
transformer. In general, the third harmonic is not used because
it can be filtered out by delta-connected windings. As a
consequence, the second and fourth harmonics are usually
preferred for inrush current detection [2-4].
Several different methods exist for detection of inrush
currents. One possibility is the use of the so-called harmonic
blocking technique. In this technique, the ratio of second
and/or fourth harmonic in the differential current of each phase
is evaluated. If this ratio exceeds a preset value, a blocking
signal is generated. So, the relay takes the decision to trip or
not to trip by combining the block signals of each phase with
the operation signal generated by the differential protection
function of each phase. There are several ways to do this
combination:

Independent blocking the blocking signal of each

phase inhibits the operation signal of this phase only
[2-4].

Crossblocking the blocking signal of any phase

inhibits all three operation units [2-4].

2-out-of-3 blocking if two blocking signals are

generated, all three operation units are inhibited [3]

Average blocking only one block signal is

generated based on the average of the harmonic content
of all three phases [3].
Another technique used to detect inrush currents is the
harmonic restraint [4]. In this technique, the percentage
restraint differential characteristic is changed according with
the second and/or fourth harmonic content by the addition of a
constant c, such as shown in equation (5) below.
I diff SLP I rest + c

Fig. 1. Percentage restraint differential characteristic

Despite the combined use of Idiff and Irest, some transient

phenomena can lead to false differential currents and the
consequent undesired operation of the protective scheme. One
of the most frequent examples of this are inrush currents
during transformer energization. It is therefore required that
differential protection relays use some technique to detect
inrush currents and prevent misoperations. Some of these
techniques are described as follows.

(5)

The constant c depends on the second and/or fourth

harmonic content. So, when the protected transformer is
energized, the percentage restraint differential characteristic is
offset, preventing the relay misoperation.
All foregoing techniques are used in this paper for
evaluating the performance of the differential protection of a
power transformer.
III. RELIABILITY OF RELAY SETTINGS
Reliability R of a device (or a system) is the probability of
this device (or this system) to perform its intended function for
a specified time under certain preset conditions [5]. The
complementary of the reliability R is the failure distribution
function F. Equation (6) shows the relationship between these
two quantities.

R (t ) = 1 F (t )

(6)

The failure distribution function F expresses the total

number of failures expected in an initial population of devices
in a certain time. Another quantity called failure rate, ,
expresses the frequency of occurrence of these failures. The
complementary of is the success rate P. Equation (7) shows
the relationship between these two quantities.
P (t ) = 1 (t )

A. Modeling of system components

(7)

The failure rate is a function of time and its most famous

time behavior is given by the bathtub curve shown in Fig. 2-A.
However, there are other kinds of failure curves as shown in
Figs. 2-B to 2-F [6].

1) System Equivalent
The system equivalent seen from the busbar at which the
transformer protection is analyzed was modeled with a
balanced three-phase voltage source behind an RL coupled
element represented in symmetrical components. The values of
the positive and zero sequence resistance and reactance used in
this study (see Table I) were obtained using the short-circuit
analysis software ANAFAS [8], which contains a complete
and accurate description of entire Brazilian power system.
TABLE I SYSTEM EQUIVALENT
R(pu)
X(pu)
Positive Sequence
2.6288
9.3253
Zero Sequence
6.5806
20.648

Fig. 2. Different kinds of failure rate curves [6].

The initial part of the bathtub curve shown in Fig. 2-A is

called Infant Mortality. It can be controlled by a procedure
called Burn-in. In this procedure, all devices are turned on
and remain in operation at the factory during a certain time.
So, all failures that are supposed to occur in the beginning of
the devices lifetime will occur before this device goes to the
final user.
The final part of the bathtub curve is related to the wear out
failures. It can be controlled by the definition of the expected
lifetime of the device. So, curves B, E and F are particular
cases of the original bathtub curve A.
The aim of this paper is to evaluate the reliability of the
transformer differential protection function during certain
conditions. This consists basically in evaluating a software
reliability. In this case, the commissioning tests may be
considered as a Burn-in procedure and the Infant
Mortality can be neglected. Since the lifetime of the
differential function is related to the lifetime of its settings, and
considering that no significant changes will occur in the
system during the lifetime of the relay, such settings will end
up being valid during a long time. In this case, the wear-out
part of the bathtub can be neglected too. As a consequence, the
failure rate will follow curve E and can be considered constant.
If is constant, then the reliability can be calculated according
to (8) [5].

R (t ) = e t

concepts of reliability. For this, a software was written in

Scilab [7] to implement a differential protection relay for time
domain transient analyses. This software is also responsible for
preparing input data cases for systematic simulation in ATP,
and for reading and analyzing the results obtained after each
simulation. Details of the implemented models are given in
Section IV-A. A description of the Monte Carlo method used
in the simulations is given in Section IV-B.

(8)

IV. DEVELOPMENTS
In this study, the differential protection of a power
transformer is tested using the Monte Carlo method and

2) Power transformer
The power transformer whose differential protection is
evaluated in this study is used at the high voltage distribution
system of COPEL, which is the major power utility company
operating in the state of Paran, south of Brazil. It is a threephase, 41 MVA, 138/34.5/13.8 kV power transformer,
supplied from the 138-kV side, which feeds distribution
networks of 34.5 kV and 13.8 kV. Since its winding
configuration is Yg-yg-, zig-zag grounding transformers are
installed at the low voltage busbar to give a reference to
ground protective relays installed at this side.
The power transformer was modeled following the same
principles used in the hybrid model available in ATP [9].
However, the following simplifications and/or improvements
were performed:

Since the studied phenomena usually do not contain

high-frequency components, capacitances and the
frequency dependence of the winding resistance were
neglected.

The full coupling matrix [A] was used instead of the

simplified matrix to take into account the difference
between zero sequence and positive sequence parameters. The importance of this assumption is illustrated in
Table II, which shows a comparison between current
values obtained for a phase-to-ground fault at the primary winding bushing using a traditional frequency
domain short-circuit analysis program (ANAFAS [8])
and ATP using both types of coupling matrix (full and
simplified). Table II shows that results obtained when
the full coupling matrix is used are closer to that obtained by ANAFAS.

TABLE II

TABLE III PARAMETERS OF THE MODELED TRANSFORMER

Primary
Secondary
Tertiary
V (kV)
138
34.5
13.8
S (MVA)
41
41
41
Conection
Yg
Yg
D
R (pu) *
0.0065
0.0046
0.0064
X1 (pu) *
0.2662
-0.0236
0.1731
X0 (pu) *
0.1336
-0.0023
0.0910
Phase Shift ()
0
30
Core loss (W)
21750

CURRENTS CONTRIBUTIONS AT PRIMARY WINDING DURING A PHASE TO

GROUND FAULT
ATP (Simplified [A])
ATP (full [A])
Phase ANAFAS
Currents (A) Currents (A) Error (%) Currents (A) Error (%)
A
679.1
420.6
38.07
678.6
0.07
B
679.1
420.6
38.07
678.5
0.09
C
679.1
421.5
37.93
679.0
0.01

To simulate turn-to-ground faults, the order of the

coupling matrix [A] was increased by splitting the
faulted winding in two.
To extrapolate the excitation curve, a fitting to the
modified Forlich equation was done, such as in the
hybrid model [9]. However, to simplify the parameter
estimation process, the empty-space inductance was
neglected and the core was represented as a triplex one.
As a consequence, the modified Forlich equation was
written as shown in (9). Using the -i points that define
the excitation curve of the transformer, it is possible to
write the overdetermined system of equations (10). The
parameters a, b and c in (9) and (10) can then be determined in a least squares sense using the
pseudoinverse method [11] as (11).

(*)Power base = 100 MVA

TABLE IV EXCITATION CURVE
12.46
13.88
14.52
1.04
1.2
2.17

V(kV)
I(A)

Energization>1_?-IAW1(A)

Energization>2_?-IBW 1(A)

15.57
7.46

Energization>3_?-ICW 1(A)

2
Magnitude (Mag)

-2

a +b i + c i
i

1
i1
1
= i2
M
1 M

n in1
11
1
2

= a i + b + c i
1
i1 2
a
1
i2 2 b

~
y = X
M
M c
1
1 in 2
1

~
~
~
~
= [( X T X ) 1 X T ] y = P y

(9)
-4
120

140

160

180
Time (ms)

200

Electrotek Concepts

(10)

220

240

TOP, The Output Processor

(a)
400
[A]
300

200

(11)

100

The magnetization branch was modeled using the

element 96 available in ATP [10]. To evaluate the
effect of hysteresis, the Hysteresis ATP routine was
used.

-100

-200

-300

-400
0,00
(file energ3.pl4; x-var t) c:X0002A-PRIA

0,02
c:X0002B-PRIB

0,04

0,06

0,08

[s]

0,10

c:X0002C-PRIC

(b)

The parameters of the transformer model are listed in Table

III. Details of the excitation curve are listed in Table IV.
To verify the transformer model, a comparison was made
between inrush currents measured during the energization of
the power transformer evaluated in this study and currents
calculated with its model. The results are shown in Fig. 3. It
must be noted that the simulated case is not identical to the
condition in which the measurements were performed because
of the difficulty of evaluating both the circuit breaker close
time and the residual flux of each phase. However, the
magnitude and shape of the obtained currents are in good
agreement with measured data. This suggests that the
simplifications adopted in the transformer model seem
reasonable.

Fig. 3. Comparison between oscillographic records obtained (a) during a real

transformer energization (currents in secondary values for a CT ratio of 600-5
A) and (b) through simulation using the proposed transformer model.

It is important to note that the aim of this study is to

develop a methodology to evaluate the reliability of the
transformer differential protection function. If necessary, the
transformer model can be improved for future studies.
3) Zig-zag Grounding Transformers
The zig-zag grounding transformers used to provide a
ground reference to the low-voltage side of the simulated
transformer were modeled as a RL coupled element in
symmetrical components. The parameters are listed in Table
V. The positive sequence resistance and reactance were set
with a very large value (106). However, the zero sequence

resistance and
specifications.

reactance were

set according COPEL

TABLE V PARAMETERS OF THE ZIG-ZAG TRANSFORMERS

R(ohms)
X(ohms)
Positive Sequence
106
106
Zero Sequence
0.20
6.78

4) Differential Protection Relay

The digital protection relay implemented in Scilab is based
on a real digital relay that can use either harmonic restraint,
independent blocking or crossblocking of second and fourth
harmonic for preventing undesired operation during
transformer energization [4]. The implemented model reads
output currents calculated in ATP using an integration step of
100 s and performs the phasor estimation at a sample rate of
64 points per 60 Hz cycle. So, the first procedure executed by
the relay model function is the interpolation of the current
input signal.
After interpolation, a low-pass filtering is done with a third
order Butterworth filter. The cutoff frequency was set to 300
Hz because the relay model requires the fourth harmonic
component (240 Hz) of the signals. The filtered signal is sent
to a 1-cycle cosine filter with 64 samples per cycle for phasor
estimation. The implemented cosine filter is similar to the one
described in [16]. Each estimated phasor is divided by the gain
of the low-pass filter at its respective frequency.
For calculating the differential and restraint currents using
(1) and (2), respectively, it is necessary to correct the angular
displacement of the input currents. This is done using the
compensation matrices defined in [4]. The differential and
restraint currents are then compared following an operation
characteristic similar to that shown at Fig. 4. If the operating
region is reached, the pick-up signal is set to 1. Otherwise, it is
set to 0. Then, this pick-up signal is integrated and if the result
reaches 2 it generates the trip signal, such as shown in Fig. 4.
If the result of this integral becomes less than zero, it is setted
to zero because it could not have a negative value.

Fig. 4. Integration procedure used by the implemented relay model.

The c constant in (5) is given by [4]

c=

100
100
I diff 2 +
I diff 4
PCT 2
PCT 4

(12)

where Idiff2 e Idiff4 are the second and the fourth harmonic
contents of the differential current. PCT2 and PCT4 are the

reference settings used by the relay. All relay settings used in

this study are listed in Table VI.
Settings
CTR1
CTR2
CTR3
TAP1
TAP2
TAP3
W1CTC
W2CTC
W3CTC
O87P
SLP1
SLP2
IRS1
U87P
PCT2
PCT4
HRSTR
IHBL

TABLE VI SETTINGS OF THE RELAY MODEL

Setting Description
Current transformer ratio of primary winding
Current transformer ratio of secondary winding
Current transformer ratio of tertiary winding
Tap value of primary winding
Tap value of secondary winding
Tap value of tertiary winding
Compensation matrix of primary winding
Compensation matrix of secondary winding
Compensation matrix of tertiary winding
Minimmun differential current
Slope of the first section
Slope of the second section
Restraint current that divides first and second section
Unrestrained differential element pickup current
Reference percentage of second harmonic content
Reference percentage of fourth harmonic content
Enables harmonic restraint technique
Enables independent harmonic blocking (if setted as
N the crossblocking method is used)

Value
60
160
400
2.91
4.36
4.36
11
11
0
0.35
30
60
6
10
20
10
Y
Y

B. Monte Carlo Method

The behavior of a differential relay during a specific
phenomenon can be defined by a binary function f(a). This
function can be set to 1 if the relay operates properly and to
zero if it fails. However, the argument a of this function is a
vector of random variables, which makes the Monte Carlo
Method particularly suitable for a systematic analysis of the
relay behavior and for an estimate of its reliability.
In this study, the reliability of the differential function
performed by the relay model is calculated for four different
events: transformer energization, external faults, faults at the
transformer bushings, and turn-to-ground faults. For each
event, it is possible to define a different set of random
variables that compose the arguments of vector a. For
instance, for an energization study, the vector a can be
represented as (13). The first element of this vector defines the
closing time of the circuit breaker pole of the phase A. The
second and the third elements define the difference between
the closing time of the circuit breaker pole of phase A and
phases B and C, respectively. The last three elements define
the residual flux of the three phases. In (13), tpre is the preenergization time and max is the peak magnetic flux. U(a,b) is
a uniform random variable, in which a and b are the minimum
and maximum values, respectively. N(, ) is normal random
variable with mean and standard deviation .
1a
U (0,1)
t pre + 60
a

a N (0, )

a = 3 = N (0, )

a
4 max U (0,1)

a
5 max U (0,1)
a

6 max U (0,1)

(13)

After defining the random variables, the next step is to

proceed with the generation of the random numbers that will
be used at each simulation. To do this, the function grand
available in Scilab is used [7].
After generating the random numbers, ATP data cards are
created for each case by a Scilab routine, such as its .bat file.
After this, the .bat file is automatically executed performing
the simulation and the .lis output file is read back by the Scilab
routine. Then, the behavior of the differential protection
function is evaluated for each case. The number of required
simulations is defined as follows: Since the function f is
binary, the mean and variance 2 of the estimator of its
performance can be calculated as (10) and (11), respectively
[12].

Finally, the value of the success rate is presented with its

error margin. In this study, a three standard deviation error
margin is used.
V. APPLICATION OF THE PROPOSED METHOD
To evaluate the proposed method, simulations were
performed for testing the reliability of the differential relay
protection of the power transformer described in Section IV-A
in four different conditions: transformer energization, external
faults, faults at the transformer bushings, and turn-to-ground
faults. The obtained results are presented in the following
sections. Details of the transformer and networks parameters,
as well as the relay basic settings are presented in Section IV.
A. Transformer energization

=p
2 =

p (1 p )
n

(14)
(15)

where p is the rate of correct operations and n is the number of

performed simulations. If the success rate P of the differential
function is calculated as the rate of correct simulations p, the
failure rate and reliability R can be calculated according to
(7) and (8).
Following (15), there is a relationship between the number
of simulations needed and the standard deviation . Thus, for a
given standard deviation, the number of needed simulations
can be calculated as
n=

p (1 p )

(16)

However, the number of simulations depends of the rate of

correct operations, and this rate changes at each simulation.
So, an iterative procedure was used to define the number of
simulations [13]. The sequence of operations used in this
procedure is shown in Fig. 5.

To evaluate the reliability of the relay protective function

during transformer energization, simulations were performed
considering three different inrush detection techniques
discussed in Section II, namely harmonic restraint,
independent harmonic blocking, and harmonic crossblocking.
Table VII lists the obtained results, where it is shown that the
best technique to prevent undesired relay operation during
transformer energization is the crossblocking. However, this
technique may block the relay incorrectly when the
energization occurs under a phase-to-ground or a phase-tophase fault due to the characteristic behavior of the healthy
phases [14].
TABLE VII
PREDICTED SUCCESS RATE FOR THREE DIFFERENT INRUSH DETECTION
TECHNIQUES IMPLEMENTED IN THE RELAY MODEL (SEE DEFINITION OF
PARAMETERS HRSTR AND IHBL IN TABLE VI)
Number of
Success Rate
Technique
HRSTR IHBL
simulations
(%)
Harmonic Restraint
Y
Y
847
90.67 3.00
Independent
N
Y
2144
68.89 3.00
Harmonic Blocking
Crossblocking
N
N
100
99.00 2.98

(*) The success rate is saturated in 100%.

The reliability of the harmonic restraint technique together

with the independent harmonic blocking for different
adjustments of second and fourth harmonic components is
evaluated in Table VIII. According to Table VIII, the best
results are obtained when the reference harmonic rate of
second and fourth harmonic is set to 15% and 10%,
respectively. It also suggests that these values should be
adjusted as small as possible, taking into account the limits of
the acquisition system (current transformers accuracy and relay
A/D converters resolution). However, additional tests have to
be performed to ensure that this set of parameters will not
block the relay incorrectly during a fault condition. It is also
worth noting that the use of the fourth harmonic content seems
a very useful feature, although many relay manufacturers do
not use it.
Fig. 5. Iterative procedure that define the number of simulations

TABLE VIII
SUCCESS RATE OBTAINED USING DIFFERENT HARMONIC REFERENCE SETTINGS
IN THE HARMONIC RESTRAINT TECHNIQUE (SEE DEFINITION OF PCT2 AND
PCT4 IN TABLE VI)
Number of
PCT2
PCT4
Success Rate (%)
Simulations
20
15
1176
86.39 3.00
20
10
847
90.67 3.00
15
10
327
96.64 3.00

B. Turn-to-ground faults
Table IX shows the success rate obtained for turn-to-ground
faults using two different minimum differential settings in a
percentage restraint differential protection scheme. In the
analysis, the same set of parameters listed in Table VI was
considered, except for parameter O87P, which varied from
0.35 to 0.5. According to Table IX, a reduction in the value of
the minimum differential current increases the reliability of the
differential function during turn-to-ground faults. However, it
increases the risk of misoperation due to noise in the
secondary circuit of current transformers. Furthermore, the
obtained values suggest that percentage restraint differential
protection is possibly not a good technique to detect turn-toground faults.

model seems to be sufficiently accurate, although improvements could be made provided additional information on the
relay algorithm were made available by its manufacturer.

VII. CONCLUSIONS
This paper presents a time-domain methodology to evaluate
the reliability of a transformer protection differential function.
To account for the random nature of many factors that affect
the protection behavior, concepts related to the Monte Carlo
method and reliability theory were used. The proposed method
seems promising in the sense that it gives to the user a more
precise idea of the expected behavior of the differential relay
under different conditions. Moreover, it allows the user to test
many different relay settings to choose the ones that are more
adequate for a given application. Future studies with more
detailed transformer and relay models, as well as with the
inclusion of current transformer models, are in due course.
VIII. REFERENCES
[1]

TABLE IX
SUCCESS RATE OBTAINED USING DIFFERENT HARMONIC REFERENCE
SETTINGS IN A PERCENTAGE RESTRAINT DIFFERENTIAL PROTECTION SCHEME
(SEE DEFINITION OF PARAMETER O87P IN TABLE VI)
O87P
Number of simulations
Success Rate (%)
0.35
1831
75.87 3.00
0.5
2116
69.61 3.00

It is important to point that the vector of random variables

for turn-to-ground faults contains five elements that define the
winding and the phase under fault, the percentage of the
winding in which the fault occur, the fault resistance and its
time instant. So, during simulations mentioned in Table IX,
many different situations were evaluated.
C. External faults and faults in the internal bushings
For external faults and internal bushing faults, the obtained
reliability was 100%. This was expected because transformer
differential protection was designed taking into account
basically these two conditions. However, it is important to note
that this result could be different if the magnetization branch of
the current transformers was taken into account during
modeling.
VI. COMPARISON WITH EXPERIMENTAL DATA
The Scilab routine stores all output files of cases classified
as unsuccessful in the form of .pl4 files. When the settings of
table VI were used, 847 cases of transformer energization were
performed, with 79 unsuccessful cases. To evaluate the
validity of the implemented relay model, the output currents
associated to such unsuccessful cases were applied into a real
digital differential relay using a power system simulator. In 67
out of these 79 cases a trip signal was sent by the relay, which
was its expected behavior. This result shows that the relay

[2]
[3]
[4]

[5]
[6]
[7]

[8]
[9]

[10]
[11]
[12]
[13]
[14]

[15]
[16]

J. R. M. S. Souza, Development of a methodology based on the Monte

Carlo method for evaluation of the reliability of differential relay
settings used in transformer protection" (in portuguese), M. Sc.
dissertation, PPGEE/UFMG - Federal University of Minas Gerais,
Brazil, 2012.
G. Ziegler, Numerical Differential Protection Principles and
Applications, Nuremberg: Publics Coorporate Publishing, 2005.
GE Multilin, GE Industrial Systems T60 Transformer Management
Relay, UR Series Instruction Manual, 2006.
Schweitzer Engineering Laboratories, Schweitzer Engineering
Laboratories SEL-387E Current Differential and Voltag Protection
Relay, Instruction Manual, 20050614, 2005, p. 3-5.
W. R. Blischke, D. N. P. Murthy, Reliability Modeling, Prediction and
Optimization, Danvers: John Willey & Sons, 2000, p. 18.
J. Moubray, Reliability-centered Maintenance, New York: Industrial
Press, 1997, p. 12.
Introduction
to
Discrete
Probabilities
with
Scilab;
http://www.scilab.org/content/download/1105/10844/file/introdiscretepr
obas.pdf; Consortium Scilab; digiteo
ANAFAS Simultaneous Fault Analysis Program Tutorial (in
portuguese), version 6.2, June, 2012
H. K. Hydalen, B. A. Mork, F. Gonzales, D. Ischenko, N. Chiesa,
Implementation and Verification of the Hybrid Transformer Model in
ATPDraw, presented at International Conference on Power Systems
Transients IPST07, Lyon, France, June, 2007
Rule Book Alternative Transient Program, CAUE Comite
Argentino de Usuarios de EMTP-ATP
A. T. Johns, S. K. Salman, Digital Protection for Power System, Second
Edition, IEE Power Series, Peter Peregrinus Ltd., 1995.
J. M. Hammersly, D. C. Handscomb, Monte Carlo Methods, London:
Chapman and Hall, 1965.
S. M. Ross, Simulation, Fourth Edition, San Diego: Elseiver, 2006.
J. R. M. S. Souza, L. V. S. Puppi, J. N. R. Romeiro Filho, M. A. S.
Mikilita, Statistical Analysis of Inrush Detection Functions Based on
Harmonic Content (in portuguese), preseted at X STPC Technical
Seminar of Protection and Control, Recife, October, 2010
S. H. Horowitz, A. G. Phadke, Power System Relaying, Somerset:
Research Studies Press, 1993.
E. O. Schweitzer III, D. Hou, Filtering for Protective Relays;
https://www.selinc.com/literature/TechnicalPapers/; Consortium Scilab;
Schweitzer Engineering Laboratories, Inc.